Forgot your password?
typodupeerror
Hardware

The CPU: From Conception to Birth 179

Posted by michael
from the cradle-to-grave dept.
CrzyP writes "Most of us have seen flowcharts and heard lectures on how a CPU functions in a computer. What a lot of us do not know, however, is how a CPU is created. Sudhian describes the step-by-step process of how a CPU is made, from grains of sand to a wafer of circuits. Ahhh sand, the building block of life...in the tech world!"
This discussion has been archived. No new comments can be posted.

The CPU: From Conception to Birth

Comments Filter:
  • Link? (Score:2, Insightful)

    Seems broken.

    Hmmm... Sand ...
  • well.. (Score:5, Funny)

    by Anonymous Coward on Saturday November 06, 2004 @12:05AM (#10740232)

    It's slashdotted already so here's the poop:

    1 Write out chip functions.
    2 Emulate on high end computers.
    3 Tape out prototypes.
    4 Port Linux to new chip.
    5 Send SCO US$699 per core.

    • Re:well.. (Score:3, Funny)

      by master_p (608214)
      You forgot some other very important steps:

      1. design a CPU in just 5 days for a market opportunity you suddently saw (*cough* 8086 *cough*).
      2. pray that this CPU will not dominate the market since it is really crap and you have better designs anyway.
      3. watch said CPU dominate the market for the next 20 years.
      4. twist your arm to find ways to speed it up.
      5. buy and burry other much better CPU architectures (*cough* Alpha *cough*) because the beast you created does not die in any way.
  • by xmas2003 (739875) on Saturday November 06, 2004 @12:08AM (#10740240) Homepage
    Site is getting pretty doggy ... here is the obligatory link to the Google Cache [66.102.7.104]
  • Summary (Score:2, Insightful)

    by Anonymous Coward
    If you didn't already know what was in the article, you shouldn't be on Slashdot.

    Slashdot: News for dorks who try to pass off as nerds.
  • by Anonymous Coward on Saturday November 06, 2004 @12:20AM (#10740268)
    but here's the scope:

    When a daddy CPU and a mommy CPU really loves each other, they get together reeeal close and...
  • by Laser Lou (230648) on Saturday November 06, 2004 @12:20AM (#10740271)
    did computation begin at conception, or at birth?
  • DNA microarrays (Score:5, Interesting)

    by FiReaNGeL (312636) <[moc.liamtoh] [ta] [l3gnaerif]> on Saturday November 06, 2004 @12:21AM (#10740274) Homepage
    Ok... so the article is not exactly new, nor interesting, so I'm gonna talk about something related :

    DNA microarrays from Affymetrix, used to quantify gene expression, are built on a process inspired from CPU design (photolitography - read more about it here [affymetrix.com]). Chips are getting more complex with time, ala Moore Law (shrinking the probe cells to get more density); the most recent human chip harbor 1 300 000 probes representing 39000 transcripts and variants.

    So technology developed for CPU is helping to find cures for diseases, increase our knowledge of life... etc. Isn't cool?
  • by Anonymous Coward on Saturday November 06, 2004 @12:21AM (#10740276)
    It's fairly short and pretty generalized. Lots of pretty pictures though.

    A quick search on Google ("silicon fabrication introduction") turns up arguably better links.

    One from SGS Thompson [eteonline.com]
    A basic one from Intel [intel.com]
    From Bell Labs [bell-labs.com]

    And there are plenty more.
  • So dull... (Score:2, Informative)

    by TiMac (621390)
    The author blurs sentences together like a 6 year old child might, using the same sentence construction over and over sometimes. It's certainly not a FUN read, but has some interesting info in it.
  • So what came first? The CPU? Or the computer that built the CPU?
  • by Dorsai65 (804760) <dkmerriman@g m a i l .com> on Saturday November 06, 2004 @12:29AM (#10740303) Homepage Journal
    Use the silicon for processors, or implants... processors, or implants...
  • by jgardn (539054) <jgardn@alumni.washington.edu> on Saturday November 06, 2004 @12:30AM (#10740306) Homepage Journal
    I have often wondered about what exactly goes into the technology we take for granted.

    The thought experiment I perform is to imagine what it would take to get the end product from absolutely nothing except the stuff around you found naturally. Working in the basement of the University of Washington physics laboratory, I often wondered how someone would build a milling machine or an industrial lathe. You can cut wood with rudimentary tools, and making crude iron or steel tools isn't too complicated, but how would construct a precise machine with all the guages and dials and electric motors and so on?

    It sure brings me to a realization of just how far we have come from slogging about in mud and eating rats like we did in the dark ages. Our world is so complicated that no one person can understand more than a small fraction of it. Everyone is a specialist of one sort of another, even the garbage collectors and sewage system maintainers. Every generation of worker brings ingenuity to the job, and bit by bit their job becomes more and more complicated yet efficient.

    Soon, will we each have a small chunk of humanity's experience in our skulls? Will we rule an insanely complicated world governed by machines and processes no one can fully understand? Or have we already come to that point?
    • by abulafia (7826) on Saturday November 06, 2004 @12:55AM (#10740388)
      This is an extremely interesting thought to me, and I've been playing with it mentally for a while now. What happens at the limits of optimization?

      Vinge [sdsu.edu], and others, have played with this concept in a sci-fi arena, but I wonder - what happens when, to take your example, garbage men hit the wall on efficiency at disposing garbage? (This implies the whole supply chain - or perhaps I should say the removal chain - of garbage mitigation specialists hitting a limit, including recyclers, dumpers, shippers, lobbyists, specialist accountants, etc.) Inputs to the garbage industry will likely be still capable of increasing demand (or, again oddly for this example, an aspect of supply), so economics start kicking in, raising costs of disposal. With garbage, we're seeing the start of this already, and in some extreme cases, lots of noise (a certain mountain in Navada [yuccamountain.org], for instance).

      This has, in turn, second order effects for lots of other industries and people, and almost nobody understands the problem, other than the people who are the maxed out specialists, for a given social, technological and economic milleu. Problems, solutions and examples of poor communication and scams start to multiply.

      It is fun stuff to think about, especially because I think we're getting a little close in certain areas. I hope to have a paper out on this soonish.

      • One of the problems with garbage disposal is that most garbage is the result of inefficient use of resources. Sometimes people can use the "dumps" to get extra resources that the original people who threw the stuff away never thought about.

        A classic example of this was the gold mines in Virginia City, Nevada. For about 10-15 years miners spent quite a bit of time (and very dangerous effort) trying to extract gold out of the mountains around the city. They started to dump the tailings from the gold minin
        • There certainly is a small amount of "waste" that somehow has to be dealt with, but the point I'm making here is that there is considerable room for improvement, and we are no where near the limits you seem to be implying.

          I may have been misleading, or you may have misread - I'm not interested in waste disposal, per se - that was just an example picked up from the parent post, used to talk about a different topic. My little kick is about systemic interactions when progress in one or more elements of the

    • "You can cut wood with rudimentary tools, and making crude iron or steel tools isn't too complicated, but how would construct a precise machine with all the guages and dials and electric motors and so on?"

      You can attach a wire, battery, and light and have a simple circuit. Even making an AND or OR circuit isn't real complicated but how do you construct billions of these circuits in order to type this simple comment.
      I think each of us already has small chunk of humanity's experience in our skulls.

    • by pipingguy (566974) on Saturday November 06, 2004 @01:47AM (#10740510) Homepage

      It sure brings me to a realization of just how far we have come from slogging about in mud and eating rats like we did in the dark ages.

      "Oh, Denis, there's some lovely filth down here!"

      It only takes a few days in complete, freezing electrical darkness to realize how dependent and utterly helpless big cities can be (and therefore its citizens) without technology [imiuru.com].

      Luckily in 1998 there were lots of people less troubled to help us out, and people mobilized from everywhere possible.
    • Yeah one interesting question is how did they get the first perfectly rectagular, cylindrical or regular shaped object. It is not trivial to get it without a mold. It must somewhat involve melting and hardening something to get a strait surface then building on that surface. Grinding, turning etc...
      • It's pulled from the melt more or less in a long cylinder. As said in the article, the first seed is put into the melt to create the orientation of the crystal, then slowly turned and pulled upwards. The eventual ingot is then ground into a perfect cylinder.
    • by Animats (122034) on Saturday November 06, 2004 @02:15AM (#10740560) Homepage
      There's a classic set of five books, Build Your Own Metalworking Shop from Scrap [lindsaybks.com], by Dave Gingery, written in the 1970s. This set covers how to bootstrap up a machine shop starting from very little.

      Step one is to make a charcoal foundry, starting with a pail, fire clay, and a steel pipe. With this you can cast parts. You hand-carve wooden masters, make sand moulds, and pour molten metal into them.

      Once you can cast, the next step is to build a lathe - the simplest machine tool. You'd probably have to make a very crude lathe first, but once you have even a crude lathe, you can make round things. Then you can make a better lathe.

      The next tool is a shaper, or planer, which allows you to make flat things. You're now up to the machining technology of 1850 or so, and can make small steam engines. Take a look at a steam locomotive. It's all castings with a little finish machining. All the finish machining is either lathe or planer work - there are no milled parts with complex surfaces.

      The other early power tool, not mentioned in Gingery, is a steam hammer. You don't need that for small work, but the steam hammer is the tool that made it possible to make stuff too big to hammer out by hand. Watt's factory had a steam hammer by 1810 or so.

      Once you have the lathe and planer, you can build, with difficulty, a milling machine. Once you have a milling machine, you can build more milling machines without too much trouble. And you can build a better mill than the one you've got.

      Once you have a good mill, you can make almost anything makeable in metal.

      People have built machine tools from these books, so it's quite possible.

      • by Bender_ (179208) on Saturday November 06, 2004 @05:17AM (#10740941) Journal
        The same published also has another interesting book:
        Instruments of Amplification [lindsaybks.com] that describes how to make your own electronic and electromechanical amplifiers from scratch. Great addition if you have to restart civilization on your own!
      • by josecanuc (91) on Saturday November 06, 2004 @09:29AM (#10741334) Homepage Journal
        I've also thought about this and when it comes down to it, you come to a point where you just need lots and lots of labor.

        Following Gingery's book, you can create nearly anything. However, where are you going to get the metal from which to create these works of art/machinery? You have to find and dig ore and refine it into metals. What do you start digging with?

        I think the original bootstrap for metal (used for work, not money) was copper found in nuggets. These days it's much harder to find natural nuggets of metal -- everyone who came before has already found them!

        So you need to dig with wood, stone, and flesh tools. Find enough ore to make a shovel's worth of metal. Grind a large stone into a bowl and melt the ore. Hot fires can be created with coal and hollow reeds blowing air into them -- make sure you have plenty! Once your ore is melted, drain off the top stuff and you're left with the metal. A shovel can be hammered out of your ingot with a stone, so that's essentially the starting point of your metal tools.

        Using your more efficient metal shovel, dig more ore -- make more shovels -- find friends to help dig.

        Now, let's say you've obtained enough metal to build your lathe. How do you get it to turn? Steam engine? Nope, you don't have a running lathe yet to help you build one! One could create a large, cast metal flywheel and have your friends (and how gracious you must be to have friends like this!!) keep it going -- that will give you enough power (or rather momentum, stored energy from your friends) to turn metal on your lathe. Your two choices of high-density material are metal and stone. If you choose a metal flywheel, you and your friends have got a lot more digging to do! If you choose stone, you have to cast a few stoneworking tools, but that's probably easier than digging enough ore for 100 lbs of metal.

        Gingery's book just has you go find/buy an electric motor. ;-)

        In other words, if you *really* want to re-enact the industrial revolution, you need to be patient and have plenty of labor. The key is all in the raw materials and the labor to extract it.
        • When you are talking about where to get some extra watts of power, there are a couple of source to consider:
          • Horsepower - I'm not kidding here either. This is why the term is still used for power measurements, because it was commonly used in the past. The neat thing about horses is that horse begat more horses, and all you need is a little grass and water. Oats and salt help to make healthier horses, as does good vetinary care, but that is just a refinement of the basics. If you are talking trying to
        • I think the original bootstrap for metal (used for work, not money) was copper found in nuggets. These days it's much harder to find natural nuggets of metal -- everyone who came before has already found them!

          If you have to rebuild society because something really bad has happened, you could simply get metal from derelict machinery. After all, those nuggets had to go somewhere, it's not like they went to a big nuclear reactor and were converted into hydrogen.

          If you're just building a blacksmith shop ju

    • sorry but what your saying sounds like alot of fluff, fluff that can destroy the world. I believe everything is simple, and simple to understand, its just a matter of the person, you, me, in interpreting it and being able to express it. If you keep telling yourself its complicated as fuck, your going to make it complicated and it's going to be complicated.

      What the world has become is a collection of redundancys, obfuscation. And to top it off, a bunch of arrogant people who dont want to take the time
    • Every generation of worker brings ingenuity to the job, and bit by bit their job becomes more and more complicated yet efficient.

      In a way, nature does the same thing on an admittedly much more basic level...Fusion...Its possible that the universe it'self is becoming more complicated.
  • Isn't that supposed to be a bad thing?
  • Ahhh sand, the building block of life...

    Huh?

  • I picked this up on a visit to a Nikon factory in Japan where they make the Step & Repeat machines that make the chips and fabs as well as the precision testing equipment. To give you an idea of the optical precision required by the Step & Repeat machines that make the Fabs and chips consider this... The accuracy is equivalent to cutting all but 1 blade of grass on an entire football field to the exact same length... Prett amazing stuff.
  • by da3dAlus (20553) <dustin.grau@gmailTOKYO.com minus city> on Saturday November 06, 2004 @01:38AM (#10740487) Homepage Journal
    Ahhh sand, the building block of life...

    [an old convict and H.I. lying on their prison bunks, passing the time]
    Ear-Bending Cellmate : ...and when there was no meat, we ate fowl and when there was no fowl, we ate crawdad and when there was no crawdad to be found, we ate sand.
    H.I. : You ate what?
    Ear-Bending Cellmate : We ate sand.
    [pause]
    H.I. : You ate SAND?
    Ear-Bending Cellmate : That's right!
  • by Fiz Ocelot (642698) <baelzharon@gmai[ ]om ['l.c' in gap]> on Saturday November 06, 2004 @01:44AM (#10740497)
    I've seen a lot of cores and it seems that most of them are of a dull or silvery color; but some are more of a green/amber shiney look. So what explains that exactly? Nothing at all?

    About a year ago I bought a couple xp 1700s that overclocked amazingly high, obviously a high quality processor set aside for selling in the lower end market. It also was the green/amber shiney color.

    • I think the colours that you see are from light refraction. Due to the size of the components involved on the die it'll tend to refract the light.

      Remember, that's one of the big problems of getting smaller than 90nm, the fact that at that wavelength things like glass are rather opaque (hence the need to use quartz).

      The chips which appear dull are either using much larger structures or might have some sort of 'protective cover' over them.

      PLD.
    • by Jay-Lo (828722) on Saturday November 06, 2004 @05:57AM (#10741007)

      The green/amber part you were looking at may have been a protective coating applied when the microprocessor was packaged. Regardless, microfabricated chips can indeed be technicolored marvels.

      Most materials used in microfabrication are either transparent (insulating layers) or grey (metallization), but resulting devices can appear coloured due to optical interference [fsu.edu]. Colours present in structures of a microfabricated device are related to the thickness and composition of the patterned thin-film coatings that form the device. For a single thin film, thickness can be determined from, for example, the Michel-Lévy interference colour chart [microscopyu.com] if the birefringence of the thin film material is known. Variations in colour across a film indicate non-uniform thickness. The colour resulting from several layers of patterned thin-films is more complex to predict, but the same basic principles apply.

    • As someone already suggested, the color is due to refraction through a thickness of silicon nitride passivation and silicon dioxide interlevel dielectrics on the die. The thickness varies with the process. I've delayered many a die for failure analysis and as you strip them down the features change shade. A die totally stripped of oxide is very hard to navigate under a microscope, as its becomes very uniform and featureless. (The opposite is true under an electron microscope -- there, the topography is
  • by Moos3d (824698) on Saturday November 06, 2004 @01:45AM (#10740500)
    Now all Intel needs is stores where you can watch the chips being made. Like a Krispy Kreme!
  • Did anybody else's butthole start to hurt when they saw the Crystal Silicon Ingot [sudhian.com]?
  • by taped2thedesk (614051) on Saturday November 06, 2004 @03:06AM (#10740648)
    Here [embedded.com] is a slightly better written article on the same topic...
  • Indeed a very impressive technology, getting better and faster by every geometry.
    As a digital designer I can't help to point out that the man time invested in an ASIC these days is an order of magnitude of what it takes to build a single chip. TSMC can put a chip out in two weeks.
    Of course I'm not taking in consideration the time taken to prepare the fab to be ready for first production, but when you and your team of 10 work tirelessly for a year, two weeks turnaround time always seems amazing.

    -P@
    • TSMC can put a chip out in two weeks.

      I doubt that, unless it is a gate array.
    • And you should also keep in mind that they have done their research and engineering work already when optimizing their processes.

      If your design is flawless and they have masks. The job to fabricate the chip can be done in two weeks i think.

      Fabricating a chip with already optimized processes should be as easy as printing our design from a printer.
  • Lame... (Score:3, Informative)

    by sharkb8 (723587) on Saturday November 06, 2004 @04:38AM (#10740850)
    This is one of the lamest, most oversimplified explanations I've seen in a long time. I think I read this in high school physics.

    For example, sand is not melted in a quartz bucket to make an ingot. Sand is Si02, or quartz. THe bucket would melt, and you;'d have an ingot full of Si and 02. Sand is made into gaseous silcon, called silane gas, which is then allowed to crystallize into a solid, chunks of which are melted in a quartz bucket.
    • Actually there was writen that sand is chemically purified to electronic grade silicon, which is pretty much correct. And that electronic grade silicon is melted in a quartz bugget.
  • From the article: "As you can guess, holes don't conduct electricity very well."

    Yes they do, halve of the transistors in the CPU rely on this fact. I know what you tried to say, but mind your words..
    • Yes, I thought about that after. "Voids" would have been a better word than "holes", but the target audience who knows nothing about semiconductor physics would never pick that up. I would have been a little more specific though, had I known this was going to end up here where people actually understand what a transistor is.
  • Sometimes you wish they'd used a condom!
  • I know a lot about IC manufacture, but I keep reading articles like this one in the vain hope that one will go into enough detail to answer this:

    I've got a sample 100mm wafer on my desk with several hunderd ICs of some sort arranged in a grid on it. The ICs are only 4mm x 4mm, but the distance between them is about 0.1mm.

    What sort of cutting device is used to chop these 4x4 squares out of the die without messing up the adjacent ones?

    This wafer isn't special in any way and I'm sure other wafers would have

  • by ricky-road-flats (770129) on Saturday November 06, 2004 @08:03AM (#10741176)
    ...was a chip fab. It had just opened, everything was shiny and new, and the work I was doing meant I got to go to every department, every part of the plant. Siemens Electronics (now Infineon) ran it.

    It was like a geek's heaven inside. Everything was the best, new and working just right. They spent something like 1.5 billion pounds ($3 billion US) on the place. Hell, even the coffee machines were wonderful.

    Inside the (huge) clean room was best - fully automated monorails all over the ceiling, carrying pods of wafers around, for instance. Row upon row of ovens with pure oxygen atmospheres at several hundred degrees C, implanters using silly amount of electricity (and huge copper hooks to remove people stuck), and incredibly dangerous chemicals being piped all over (including the very scary HF - 'If it leaks near you, there's no point in running').

    Wonderful stuff. It was all incredibly interesting, to see all the processes that went into making (relatively simple) RAM chips.

    Shame the arse fell out of the DRAM market in 1999, meaning they closed the place. Atmel are using it now.

    • What's HF? It sounds neat :)
      • HF = hydrofluoric acid
      • Its the evil thing in a chipfab. Everybody talks about HF and is afraid of coming near or getting exposed. A chemist would probably say: "Dont touch", thats it.

        There are many other dangerous substances in a chip fab like silan, arsin, phosphin, chlourtriflouride (now thats nasty). But all over all the amount is pretty low and everything is sealed of insanely well. It is much more dangerous to work in a chemical plant.
      • I used to work at Lucent's Micro fab in Allentown PA. The supply lines for the "scary stuff" were all encased in coaxial lines filled with inert gas at higher than atmospheric pressure, so that you'd have to breach both lines to have an "incident."

        I too have heard that it's the most evil 30 seconds of life that you'll ever finish with, but of course there were never any problems with that crap.

        They had enough problems selling enough chips to keep me employed, and in that they failed miserably.
    • (and huge copper hooks to remove people stuck)

      Huh?
  • Wow (Score:1, Offtopic)

    by dafunn (114459)
    Maybe it's the vodka talking, but I started to read the article and, about three paragraphs in, I realized just how much I didn't care.
  • Some mistakes... (Score:3, Informative)

    by curious.corn (167387) on Saturday November 06, 2004 @08:30AM (#10741220)

    Although it's a neat effort to explain some engineering & physics to the avg case modder running XP & windowblinds (;-)) there's an initial nasty mistake:

    The new wafers are then taken and doped appropriately for the type of transistors that will be made out of them. Doping amounts to depositing other elements into the space between silicon atoms. This is what causes silicon to be the "semiconductor" that it is. Transistors today are made from "CMOS" technology, or Complementary Metal Oxide Semiconductors. Complementary means the interaction of "n" and "p" MOS

    No, no... doping is about getting impurities inside the Si lattice substituting some of the Si atoms. The whole concept is: electron energy levels of a single atom becoming thick bands for hoards of electrons to fly within; if the next band is empty & close enough to the last full band you have an "intrinsic" semic. Doping the crystal means to get other atoms (P) into the lattice so that their electrons are weakly tied and readily bumped into the conduction band (@ room temp); or you plug greedy B into the lattice so that it grabs an e- all for itself leaving some other Si without and a roaming Hole inside the last full band...Leaving doping atoms wedged inside the lattice without participating to the whole electron/lattice exchange doesn't do anything good, perhaps it just deforms the reticle creating all sorts of defects & a useless brick of solid sand

    Overall this article lacks a lot of geek factor... there's so many "cool" catchy words and processes like Silicon Over Insulator, Damascene Process, dovetail prevention, SiN and SuperK dielectric... bah, it could have been a LOT better... have a look in ars [arstechnica.com]
    • here's so many "cool" catchy words and processes like Silicon Over Insulator, Damascene Process, dovetail prevention, SiN and SuperK dielectric

      Nice buzzword dropping, but next time at least do it right - I count at least three mistakes.
      • such as? I've never been an ace in chemistry so I concede SiN, super-k could be high-k but what the hell... dovetail? well I remember some comments about the tapering edges of thin gate oxide features... BTW... I wouldn't mind some links; chip manufacture has always fascinated me.

        • Well, its "Silicon On Insulator" (SOI), Si3N4 (whatever may be exciting about it) and high-k. I also do not know about dovetail, but it may be related to problems with void filling in interconnects as you get structures similar to a dovetail there. The phenomenon you are referring to is probably the "birds peak", common in LOCOS formation.

          Sorry, I do not have any links. I'd suggest to look for some books about semiconductor prozessing e.g. S.M.Sze or Madou.
  • First the misleading part:

    CrzyP writes "Most of us have seen flowcharts and heard lectures on how a CPU functions in a computer. What a lot of us do not know, however, is how a CPU is created.

    I swear I envisioned decisions of how many registers to do what, what the instruction set should include, pipelining, hardwired vs. microprogramming, etc. Insteresting Stuff, at least to this nerd.

    BUT NOOOOOO, it's about:

    Sudhian describes the step-by-step process of how a CPU is made, from grains of sand to a wa
  • What would have been far more interesting would be to skim the surface of the actual design and layout process of the chip, the software used that actually builds the layout and chip design, and the uber-geeks that actually simulate the design, identify bottlenecks or problems in the design, and hand-optimize the chip design before committing it to the mask process ("tape"), and the tools and techniques they use to do this.

    It's a lot more than a SPICE simulation, I think...

...when fits of creativity run strong, more than one programmer or writer has been known to abandon the desktop for the more spacious floor. - Fred Brooks, Jr.

Working...